Skip to content

CHANGELOG.md

Latest commit

 

History

History
511 lines (435 loc) · 52.6 KB

CHANGELOG.md

File metadata and controls

511 lines (435 loc) · 52.6 KB
Raw

Changelog

[Unreleased]

v4.4 orchestrator-engine items continue below under their original "v4.4, orchestrator engine" headings. They are in-flight and will roll into the v4.4 entry when that release lands.

[4.3.0] - 2026-04-20

Graduated (residential_multifamily)

  • residential_multifamily subsystem: status: beta_rc (v0.6.0) → status: stable_pending_shakedown (v1.0.0-rc1). Code complete, refusal-on-missing-input contracts active; subsystem awaits the first operator shakedown log before graduation to status: stable. SKILL.md Release maturity section now carries a Stable, awaiting shakedown banner instead of the PREVIEW / STAGING stamp — per docs/PREVIEW_MODE.md, shakedown status does not layer a second final-marked refusal on top of the existing contract machinery.

Closed (tailoring Pass 2 — Objective 4)

Four items previously listed as Not implemented in docs/tailoring_capability_matrix.md are all Implemented and tested on main:

  • Approval-floor refusal (guard_approval_floor_not_lowered): tailoring cannot lower an approval threshold below the canonical floor in _core/approval_matrix.md. ApprovalFloorGuardTests (3 tests).
  • Canonical-definition redefinition refusal (guard_no_canonical_redefinition): any overlay write targeting a frozen metric field produces a REDEFINITION_ATTEMPT. CanonicalRedefinitionGuardTests (2 tests).
  • Preview-bundle YAML emission (emit_preview_bundle): writes tailoring/sessions/<org_id>/<session_id>__preview.yaml with session / diff_summary / diff_entries / proposed_overlay / guard_violations sections. PreviewBundleTests.
  • Missing-doc blocker (guard_missing_doc_blockers): refuses to render a preview when any question-bank trigger references a doc_catalog.yaml slug with no entry. MissingDocBlockerGuardTests (2 tests).

Capability-matrix doc now reflects Obj 4 closure; open follow-ups (legacy-bank retirement, role-based bank filtering) are tracked out of Obj 4.

Added (preview-mode gate)

  • tests/test_preview_mode_gate.py (repo-wide): walks every src/skills/**/SKILL.md, admits the three preview sub-statuses (experimental, beta_rc, stable_pending_shakedown), enforces the required Release maturity markers, and fails on banner-text drift per sub-status.
  • src/skills/residential_multifamily/tests/test_preview_mode_gate.py (subsystem): extended to admit stable_pending_shakedown, pin the Stable, awaiting shakedown banner, and refuse silent re-import of the PREVIEW / STAGING stamp language in a shakedown-status skill.

Added (portable ZIP structural smoke)

  • tests/install_smoke/test_portable_zip.py: structural smoke on dist/cre-skills-portable.zip. Validates ZIP opens cleanly, the skills/ tree mirrors src/skills/ minus the portable-excluded residential_multifamily subsystem, each extracted SKILL.md carries the portable frontmatter contract (name + description; slug/status/version are stripped by design per config/targets/portable.yaml), and no MCP server / orchestrator / Python calculator leaks into the artifact.
  • .github/workflows/portable-zip-smoke.yml: pull_request (relevant packaging paths) + workflow_dispatch. Builds the portable target, packages the ZIP, runs the smoke test.
  • Cross-runtime invocation on Codex / Gemini / Grok / Manus remains an explicit xfail — structural != cross-runtime.

Added (canonical Desktop marketplace caveat)

  • docs/WHAT-TO-USE-WHEN.md is the source of truth; the caveat paragraph is wrapped in <!-- CANONICAL-CAVEAT:desktop-marketplace START/END --> sentinels.
  • Duplicated verbatim into README.md, docs/INSTALL.md, docs/install-guide.md, docs/install-desktop.md, docs/install-cowork.md.
  • tests/test_release_version_parity.py gains test_canonical_caveat_source_is_nonempty and parameterized test_canonical_caveat_duplicated_verbatim — byte-level drift fails CI.
  • docs/INSTALL.md adds an explicit FAQ entry explaining why Chat tab's "Add marketplace" rejects the repo URL and what the correct install path is per surface.

Changed (v4.3.0)

  • docs/PREVIEW_MODE.md: defines stable_pending_shakedown as a sub-status. Tabulates the four sub-statuses (experimental / beta_rc / stable_pending_shakedown / stable) with banner + final-marked-refusal semantics per status. Documents the three-step graduation path.
  • docs/ROADMAP.md: current-release block flips to v4.3.0. v4.3 "Near-term hardening" items are closed; remaining items moved to a post-v4.3 open-items block. Tailoring Pass 2 Obj 4 no longer listed as open.
  • docs/tailoring_capability_matrix.md: "Deliberate next steps" block rewritten to reflect Obj 4 closure; only legacy-bank retirement and role-based bank filtering remain.
  • docs/install_smoke_test_matrix.md: Portable ZIP "Fresh install" cell flipped from gap to structural, with cross-runtime invocation explicitly still listed as a gap.
  • README.md: Release Maturity table, Known Limitations, What's New section all updated for v4.3.0 and RMF graduation.
  • .claude-plugin/plugin.json version 4.2.0 → 4.3.0.
  • .claude-plugin/marketplace.json plugin version 4.2.0 → 4.3.0.
  • src/catalog/catalog.yaml plugin_version 4.2.0 → 4.3.0.
  • Installer scripts (scripts/Install.ps1, scripts/install.sh, Install.command) fallback version strings and banners bumped to 4.3.0.
  • Docs (docs/INSTALL.md, docs/install-guide.md, docs/install-desktop.md, docs/install-cowork.md) DMG/EXE filenames and version strings bumped to v4.3.0.
  • src/hooks/telemetry-init.mjs default-config version and upgrade-backfill threshold bumped to 4.3.0.

Removed (v4.3.0)

  • README.md and docs/install_smoke_test_matrix.md no longer frame the residential_multifamily subsystem as unfinished near-term hardening — the pass-2 deferred items all close in this release.

Historical: v4.4, orchestrator engine (in-flight)

Fixed (v4.4, orchestrator engine)

  • Executor summary now prints a "paused, awaiting human approval" line for AWAITING_APPROVAL pipelines instead of falling through to the KILL-style "Pipeline terminated. Deal is not viable at current terms." message. Exit code (0) and pipeline verdict were already correct; only the summary text was mis-branched.

Added (v4.4, orchestrator engine — PR C of 3)

  • Calculator bridge (design doc section 3). New src/orchestrators/engine/calculator-bridge.mjs spawns scripts/calculator-invoker.py on demand to resolve phase tool_calls. Surfaces the invoker's structured error envelopes (validation_errors, unknown-calculator error field) via a typed CalculatorBridgeError before falling through to a generic exit-code error.
  • Executor now honors tool_calls declared on a phase: invokes each, logs per-call success/failure to stdout, and attaches parsed results to the phase result under a tool_results key (keyed by the as alias or tool slug). Emits calculator_invoked / calculator_failed audit events when running under --deal-id.
  • New --config-path <path> flag on the executor lets test fixtures and ad-hoc pipelines bypass the registry lookup. --pipeline is now optional when --config-path is set (pipeline id derived from orchestratorId).
  • tests/test_orchestrator_calculator_bridge.py (4 tests). A tiny Node driver (tests/harness_calculator_bridge.mjs) exercises the bridge directly for success, unknown-slug, and validation-error paths; an E2E test wires a minimal fixture pipeline through the executor and asserts calculator_invoked in the audit log.

Added (v4.4, orchestrator engine — PR B of 3)

  • Variant loader (design doc section 4). New src/orchestrators/engine/variant-loader.mjs resolves src/orchestrators/configs/<pipeline>/variants/<variant>/phases.json and overlays it onto the base config. Applies phase_weight_overrides and merges added_approval_gates into the target phase's approval_gates array. Unknown phase-id overrides raise — authoring bugs fail loudly, not silently. Unknown variant slug (no directory) is tolerated with a log line.
  • Typed approval gates (design doc section 2). New src/orchestrators/engine/approval-gate.mjs evaluates any gate declared on the current phase against dealState.approval_gate_log. First encounter opens the gate as pending and fires gate_opened audit event. Any gate with status pending|denied|expired blocks the phase with verdict AWAITING_APPROVAL; agents do not dispatch. Executor exits 0 on AWAITING_APPROVAL (pipeline paused, not killed).
  • Out-of-band gate clearance. New CLI src/orchestrators/engine/approve-gate.mjs flips a gate record to approved, approved_with_conditions, denied, or expired, persists state, and appends the matching audit event (gate_approved, gate_approved_with_conditions, gate_denied, gate_expired). Kept separate from executor.mjs so the executor never approves its own gates.
  • Executor applies the variant overlay after config load when --strategy matches a variant directory. Pipeline-level verdict vocabulary extended with AWAITING_APPROVAL.
  • tests/test_orchestrator_gates_variants.py (3 tests). Covers weight override (underwriting 0.2 → 0.3 in dry run), unknown variant tolerated, and the full block→approve→resume→complete cycle with audit append-only verification across the approval.

Added (v4.4, orchestrator engine — PR A of 3)

  • Persistent deal-scoped state (design doc section 1). New src/orchestrators/engine/deal-state.mjs loads, writes (atomic via .tmp+rename), and mutates <home>/.claude/cre-skills/deals/<deal_id>/state.json against the schema shipped in the v4.4 scaffold. Executor gains --deal-id <id>: first run initializes state, subsequent runs resume from verdicts_by_phase and skip already-resolved phases. Mismatched --pipeline + existing deal refuses (exit 3).
  • Human-in-the-loop audit log (design doc section 5). New src/orchestrators/engine/audit-log.mjs appends newline-delimited JSON events to <dealDir>/audit_log.jsonl. Append-only invariant: never reads-and-rewrites. Every orchestrator event (pipeline_started, phase_started/completed, pipeline_halted/completed, resume_started) carries timestamp, event, actor, deal_id, run_id.
  • tests/test_orchestrator_deal_state.py (5 tests) covers roundtrip, resume, pipeline-mismatch refusal, audit append-only (SHA-256 prefix check), and back-compat (omitting --deal-id creates no deal dir).
  • Omitting --deal-id keeps the pre-v4.4 ephemeral session flow unchanged — no existing behavior regresses.

[4.2.0] - 2026-04-16

Added (plugin v4.2.0 — hardening pass 2 close)

  • Sealed-close gating (Obj 5). New _core/schemas/period_seal.yaml canonicalizes close_status ordering (draft < soft_close < hard_close < locked) and the as_of / close_lock_timestamp / budget_version / reforecast_version fields. _core/final_marked_workflows.yaml gains a period_grade_workflows registry enumerating six slugs with their minimum close-status floor. Six workflow reference_manifest.yaml files (executive_operating_summary_generation, quarterly_portfolio_review, monthly_property_operating_review, monthly_asset_management_review, reforecast, budget_build) declare required_period_seal. tests/test_period_seal_gating.py (5 tests) enforces the contract.
  • Finance placeholder scanner (Obj 6). New _core/reference_data_integrity.md documents the rule: placeholder tokens (TBD, TODO, FIXME, XXX, PLACEHOLDER, TKTK) in any reference CSV read by a final-marked workflow require an explicit placeholder label. tests/test_finance_placeholder_scanner.py (4 tests) runs on every CSV read by the final-marked manifests.
  • Executive output contract (Obj 8). New _core/executive_output_contract.md defines verdict-first structure, source-class labels ([operator] / [derived] / [benchmark] / [overlay] / [placeholder]), and refusal-artifact shape for final-marked output. Four final-marked SKILL.md files (executive operating summary, IC prep, quarterly portfolio review, executive pipeline summary) reference the contract; the canonical example (executive_operating_summary_generation/examples/ex01_*.md) demonstrates the full pattern. tests/test_executive_output_contract.py (4 tests) enforces the references and the canonical example shape.
  • _core/schemas/reference_manifest.yaml extended to admit required_period_seal as a typed top-level field.

Changed (plugin v4.2.0)

  • residential_multifamily subsystem: status: draft → status: beta_rc, version 0.5.0 → 0.6.0.
  • .claude-plugin/plugin.json version 4.1.2 → 4.2.0.
  • .claude-plugin/marketplace.json plugin version 4.1.2 → 4.2.0.
  • src/catalog/catalog.yaml plugin_version 4.1.2 → 4.2.0.
  • README "Release Maturity" row for residential_multifamily downgraded Experimental → Beta RC; Known Limitations refreshed to reflect period-seal and executive-output-contract coverage.
  • Implementation hardening Obj 5 / 6 / 8 flipped from deferred pass 2 to done (pass 2); test totals ratcheted up.
  • Installer scripts (scripts/Install.ps1, scripts/install.sh, Install.command) fallback version strings bumped to 4.2.0.
  • Docs (docs/INSTALL.md, docs/install-guide.md, docs/install-desktop.md, CONTRIBUTING.md, PRIVACY.md) version banners and examples bumped to v4.2.0.
  • src/hooks/telemetry-init.mjs default-config version and upgrade-backfill threshold bumped to 4.2.0.

Fixed (plugin v4.2.0)

  • tests/test_catalog_claim_integrity.py::test_build_artifact_hook_prompts_match_catalog now passes vacuously when the builds/ directory is absent (fresh clones and CI). Previously asserted presence unconditionally, failing the test outside developer environments.
  • .gitleaks.toml allowlists src/skills/*/workflows/*/tools/test_*.py so deliberate fake-secret fixtures in TUI scrubbing tests do not register as real secrets.

Removed (plugin v4.2.0)

  • Stale docs/plans/build-system-refactor.md and docs/plans/residential-multifamily-refinement-2026-04-15.md (completed-plan docs per release-hygiene validator).

Security notes (plugin v4.2.0)

  • All 4 hardening-pass-2 commits + 2 release-hygiene fixup commits signed via 1Password SSH (ED25519).
  • Gitleaks, semgrep-cloud-platform/scan, secrets-scan all green on PR #31 (24/24 CI checks).
  • No new dependencies added in this release.

Added (residential_multifamily v0.5.0 - wave-5 stack completion + Yardi 2026-04-15)

  • reference/connectors/adapters/yardi_multi_role/ - Yardi adapter (multi-role posture). 43 files including manifest.yaml (multi-role mapping_hints across PMS/GL/CRM/reporting feeds, 17 known_gotchas), classification_worksheet.md (4-dimension operator decision: role / access path / operating pattern / data sensitivity), bounded_assumptions.yaml (14 assumptions with lifts_when), provisional_source_contract.yaml (21 entity families, all requires_classification: true), source_contract.yaml (organized by 7 Yardi sub-systems), normalized_contract.yaml with per-role precedence, field_mapping.yaml, dq_rules.yaml (29 rules yd_ prefix), reconciliation_rules.md + reconciliation_checks.yaml (13 checks yd_recon_ prefix; AppFolio cutover, Intacct parallel, Procore handoff, Dealpath post-IC, Excel, GraySail deferred), edge_cases.md (20 cases including legacy/parallel/historical), crosswalk_additions.yaml (10 crosswalks with effective_start + survivorship_rule + per-attribute source_of_record), workflow_activation_additions.yaml (role_by_classification for 6 existing + 10 proposed workflows), 11 sample_raw/*.jsonl + 11 sample_normalized/*.jsonl, 4 runbooks (classification_path, onboarding, common_issues, migration_to_appfolio), tests/test_yardi_adapter.py (15 tests), README.md.
  • _core/ontology.md adds canonical Commitment / PurchaseCommitment object — buy-side contractual obligation with original_amount, approved_change_orders_amount, revised_amount, paid_to_date, retainage_held, balance_to_complete, status (draft|out_for_signature|executed|in_progress|complete|terminated|cancelled), funding_source, bonded, lien_waiver_required. Procore primary; Intacct posted-spend reconciliation; Yardi historical fallback. Guardrail: no executed transition without non-expired vendor insurance + bond where required.
  • 9 new workflow packs under workflows/: pipeline_review (weekly Dealpath pipeline by stage), pre_close_deal_tracking (closing-checklist + key-date countdown), investment_committee_prep (pre-IC packet assembly with sensitivity tests), acquisition_handoff (deal-close → operating asset coordination across AppFolio + Intacct + vendor master), post_ic_property_setup (placeholder property/entity creation pre-close), delivery_handoff (Procore TCO → AppFolio operating activation + dev_project_crosswalk closure), development_pipeline_tracking (weekly Procore project rollup vs baseline), lease_up_first_period (monthly ramp tracking 6-18 months post-delivery), executive_pipeline_summary (monthly board-ready rollup of pipeline + capex commitment exposure). Each pack ships SKILL.md (frontmatter + body), reference_manifest.yaml, routing.yaml, change_log.md, examples/.
  • Wave-5 fill of 7 wave-4 adapters (all reaching the 17-deliverable bar):
    • appfolio_pms/: 5 sample_raw + 5 sample_normalized JSONL files (work_orders, leads, applications, vendors, turns); dq_rules.yaml (26 rules af_ prefix); reconciliation_rules.md; reconciliation_checks.yaml (13 checks af_recon_); edge_cases.md (12 cases); crosswalk_additions.yaml; workflow_activation_additions.yaml; runbooks/appfolio_onboarding.md + appfolio_common_issues.md; tests/test_appfolio_adapter.py (26 tests).
    • sage_intacct_gl/: 5 sample_raw + 5 sample_normalized JSONL files (actual_lines, budget_lines, forecast_lines, vendors, projects); dq_rules.yaml (23 rules ic_ prefix); reconciliation_rules.md; reconciliation_checks.yaml (16 checks ic_recon_); edge_cases.md (15 cases); crosswalk_additions.yaml; workflow_activation_additions.yaml; runbooks/sage_intacct_onboarding.md + sage_intacct_common_issues.md; tests/test_sage_intacct_adapter.py (26 tests).
    • procore_construction/: 7 sample_raw + 7 sample_normalized JSONL files (projects, commitments, change_orders, draw_requests, schedule_milestones, vendors, cost_codes); dq_rules.yaml (21 rules pc_ prefix); reconciliation_rules.md (5 cross-system sections including Procore↔Yardi delivery handoff); reconciliation_checks.yaml (13 checks pc_recon_); edge_cases.md (13 cases); crosswalk_additions.yaml; workflow_activation_additions.yaml; runbooks/procore_onboarding.md + procore_common_issues.md; tests/test_procore_adapter.py (17 tests).
    • dealpath_deal_pipeline/: full source_contract.yaml, normalized_contract.yaml, field_mapping.yaml; 6 sample_raw + 6 sample_normalized JSONL (deals, assets, deal_milestones, deal_key_dates, ic_decisions, deal_team); reconciliation_rules.md; reconciliation_checks.yaml (13 checks dp_recon_); edge_cases.md (15 cases); crosswalk_additions.yaml (asset, property_master, dev_project, market crosswalks); runbooks/dealpath_onboarding.md + dealpath_common_issues.md; tests/test_dealpath_adapter.py (11 tests).
    • excel_market_surveys/: 5 new template_schemas (vendor_rate_card, turn_cost_library, capex_cost_library, schedule_assumption, market_commentary); normalized_contract.yaml (11 canonical objects with as_of_required: true); field_mapping.yaml; 5 sample_valid CSVs + 3 sample_invalid CSVs; dq_rules.yaml (21 rules ex_ prefix); reconciliation_rules.md; reconciliation_checks.yaml (12 checks ex_recon_); edge_cases.md (16 cases); crosswalk_additions.yaml; workflow_activation_additions.yaml; runbooks/excel_onboarding.md + excel_common_issues.md; tests/test_excel_adapter.py (13 tests).
    • manual_sources_expanded/: file_family_registry.yaml (12 file families); normalized_contract.yaml; 7 sample_files CSVs (TPM monthly, variance narrative, bid tab, approval matrix threshold, escalation log, payroll summary, manual property correction); reconciliation_rules.md; reconciliation_checks.yaml (15 checks ms_recon_); edge_cases.md (16 cases); crosswalk_additions.yaml; runbooks/manual_sources_onboarding.md + manual_common_issues.md; tests/test_manual_sources_adapter.py (8 tests).
    • graysail_placeholder/: runbooks/graysail_classification_path.md (operator decision tree with 8 checkpoints); dq_rules.yaml (10 deferred-mode rules gs_ prefix, all severity warning until classification closes); edge_cases.md (10 cases); workflow_activation_additions.yaml (every workflow in partial_mode_behavior: blocked_pending_classification); tests/test_graysail_adapter.py (7 tests).
  • reference/normalized/schemas/reconciliation_tolerance_band.yaml - canonical tolerance-band registry referenced by every adapter dq_rules.yaml and reconciliation_checks.yaml. ~37 named bands (close_feed_lag, capex_posting, dev_handoff_lag, ledger_drift, occupancy_drift, etc.) with default_value, severity_below_band / severity_above_band, overridable_by overlay scopes. Closes a wave-3 reference gap surfaced by wave-4 adapters.
  • pyproject.toml at repo root configures pythonpath so skill-local tests collect cleanly when pytest is invoked from project root. Closes the wave-3 conftest collection regression that forced developers to cd src/skills/residential_multifamily/tests/ before running pytest.
  • Wave-5 Yardi integration into registries: vendor_family_registry.yaml adds yardi_multi_role + yardi_voyager_classified_template; source_registry/source_registry.yaml adds 5 Yardi source records (yardi_voyager_pms_stub, yardi_voyager_gl_stub, yardi_rentcafe_stub, yardi_data_connect_stub, yardi_legacy_export_stub); master_data/ crosswalks (property_master, unit, lease, vendor_master, account, capex_project, dev_project, asset, market, submarket, resident_account) gain Yardi rows with effective_start: 2026-04-15 and provenance: wave_5_yardi_adapter.
  • _core/stack_wave4/source_of_truth_matrix.md updated: commitment row no longer placeholder (canonical object now exists in ontology); Yardi precedence is operator-classified, with five role profiles (yardi_primary_operating, yardi_primary_accounting, yardi_primary_leasing_only, yardi_legacy_historical, yardi_parallel_partial) defining how Yardi displaces or co-exists with AppFolio / Intacct / Dealpath / Procore.

Changed (residential_multifamily v0.5.0)

  • _core/ontology.md - canonical objects: 32 → 33 (commitment added). Wave-5 is otherwise additive; no canonical metric, routing axis, or workflow-pack contract was modified.
  • tests/test_source_registry.py ALLOWED_DOMAINS extended to include deal_pipeline (the wave-4 source domain that was added to the schema but not to the test guardrail).
  • Adapter inventory: 15 → 16 (yardi_multi_role added). Workflow pack count: 26 → 35 (9 wave-5 packs added). Canonical objects: 32 → 33.

Migration notes (residential_multifamily v0.5.0)

  • Wave 5 is purely additive. Wave-4 adapters that referenced commitment as a placeholder canonical now resolve against the new ontology object; no rename required.
  • Operators running pytest from project root no longer need to cd into src/skills/residential_multifamily/tests/. Existing skill-local test invocations continue to work.
  • The new Yardi adapter ships at status: stub and is gated behind yardi_multi_role/runbooks/yardi_classification_path.md. No workflow trusts Yardi data until that classification closes for the operator's environment.
  • Several tolerances cited by wave-4 adapter rules now resolve against reference/normalized/schemas/reconciliation_tolerance_band.yaml. Overlays may override per band via the overridable_by mechanism.

Added (residential_multifamily v0.4.0 - stack-specific operationalization 2026-04-15)

  • reference/connectors/adapters/appfolio_pms/ - AppFolio PMS adapter. Manifest, source_contract, normalized_contract, field_mapping, mapping_template, sample_raw/ (properties, units, leases, lease_events, tenants, charges, payments), source_registry_entry, tests. Primary source for unit roster, lease, charge, payment, work order, vendor directory, leasing funnel; secondary for property and vendor master.
  • reference/connectors/adapters/sage_intacct_gl/ - Sage Intacct GL adapter. Manifest, source_contract, normalized_contract, field_mapping, sample_raw/ (chart_of_accounts, dimensions, journal_entries), source_registry_entry, tests. Primary for budget/forecast/actuals/variance; survives AppFolio for posted actuals after close.
  • reference/connectors/adapters/procore_construction/ - Procore construction adapter. Manifest, source_contract, normalized_contract, field_mapping, mapping_template, source_registry_entry, tests. Primary for construction project, commitments (placeholder), change orders, draws, schedule milestones; reconciles to Intacct on capex postings.
  • reference/connectors/adapters/dealpath_deal_pipeline/ - Dealpath deal-pipeline adapter. Manifest, README, dq_rules, workflow_activation_additions, source_registry_entry, tests. Primary source for pre-close deal state, IC milestones, deal team assignments, deal key dates. Seeds downstream AppFolio/Procore/Intacct setup.
  • reference/connectors/adapters/excel_market_surveys/ - Excel market-survey intake framework. Manifest, README, intake_manifest, 7 template_schemas (rent_comp, concession_tracker, market_survey_workbook, analyst_benchmark_pack, staffing_benchmark, labor_rate_sheet, material_price_sheet), source_registry_entry (5 file families), tests. Treats Excel as first-class production input with provenance, staleness, reviewer attestation.
  • reference/connectors/adapters/manual_sources_expanded/ - Expanded manual-source coverage for emailed files, shared-drive submissions, operator workbooks, approval matrices, TPM submissions, bid tabs. Manifest, README, dq_rules, workflow_activation_additions, source_registry_entry (4 file families), tests.
  • reference/connectors/adapters/graysail_placeholder/ - GraySail non-destructive integration stub pending classification. Manifest, README, classification_worksheet, bounded_assumptions, provisional_source_contract, workflow_relevance_map, source_registry_entry (status: planned), tests. Workflows blocked from treating GraySail as primary while placeholder.
  • reference/connectors/deal_pipeline/ - new vendor-neutral connector domain for deal pipeline data. Manifest, schema, mapping, sample_input, README. Backward-compatible enum extension to source_registry.schema.yaml::source_domain.
  • reference/connectors/_core/stack_wave4/ - cross-cutting wave-4 design docs: source_of_truth_matrix.md (48 canonical objects with primary/secondary/fallback + disagreement resolution), lifecycle_handoffs.md (8 explicit cross-system handoffs with payloads), stack_reconciliation_matrix.md (21 stack-level reconciliations), stack_rollout_wave4.md (4 sub-waves: 4A/4B/4C/4D), third_party_manager_oversight.md (3 TPM data scenarios + confidence flags), stack_test_taxonomy.md (16 test classes), open_questions_and_risks.md (14 items including ontology extensions).
  • reference/connectors/master_data/asset_crosswalk.yaml - new crosswalk for canonical asset (Dealpath primary).
  • reference/connectors/master_data/market_crosswalk.yaml - new crosswalk reconciling Excel broker naming with AppFolio internal slugs.
  • reference/connectors/master_data/submarket_crosswalk.yaml - new crosswalk for submarket name reconciliation.
  • reference/connectors/source_registry/source_registry.yaml - 14 wave-4 source entries merged from per-adapter fragments. Total registered sources: 13 -> 27.
  • tests/test_stack_wave4.py - 69 wave-4 conformance tests (adapter directory presence, manifests carry wave_4 tag, source_registry merge integrity, GraySail status remains planned, PII classifications valid, no credential strings, deal_pipeline domain exists, cross-cutting docs present, new crosswalks present). All passing.

Changed (residential_multifamily v0.4.0)

  • reference/connectors/source_registry/source_registry.schema.yaml - source_domain enum extended to include deal_pipeline. Backward compatible.
  • Adapter inventory grew from 8 vendor-neutral stubs to 8 stubs + 7 stack-specific adapters (appfolio_pms, sage_intacct_gl, procore_construction, dealpath_deal_pipeline, excel_market_surveys, manual_sources_expanded, graysail_placeholder). Total adapter directories: 8 -> 15.
  • Wave-4 net test additions: 69 stack-specific + ~8 per adapter local tests added so far (additional adapter-local tests scheduled per _core/stack_wave4/stack_test_taxonomy.md).

Migration notes (residential_multifamily v0.4.0)

  • Wave 4 is purely additive. No canonical _core/ object, metric, workflow, or routing axis was modified. Adding the deal_pipeline source domain extends the integration layer enum; not a canonical-base change.
  • Several adapters carry incomplete content from this pass (sample data partial, some runbooks pending, reconciliation_rules.md / edge_cases.md / crosswalk_additions.yaml fragments not yet authored for every adapter). Tracked in _core/stack_wave4/open_questions_and_risks.md.
  • Surfaces canonical extensions required for full sub-wave 4B operations: (1) commitment ontology object (Procore primary entity), (2) proposed new workflows (pipeline_review, pre_close_deal_tracking, development_pipeline_tracking, acquisition_handoff, executive_pipeline_summary, investment_committee_prep, post_ic_property_setup, lease_up_first_period, delivery_handoff).
  • GraySail integration deferred behind runbooks/graysail_classification_path.md workflow; sub-wave 4C blocked until completion.

Added (residential_multifamily v0.3.0 - operationalization pass 2026-04-15)

  • reference/connectors/source_registry/ - source inventory and system coverage matrix. 13 starter records spanning all 8 source domains plus planned entries. Includes schema, vendor family hints, implementation inventory.
  • reference/connectors/master_data/ - entity crosswalk framework covering property, unit, lease, resident, vendor, account, capex project, change order, draw request, employee, dev project. Identity resolution framework with canonical/source ID separation, confidence scoring, survivorship rules, effective dating. Manual overrides and unresolved exceptions queue.
  • reference/connectors/hr_payroll/ - HR/payroll/staffing connector stub with 8 entities (employee, staffing_position, role_assignment, property_assignment, vacancy_status, payroll_line, overtime_line, employee_vs_contractor_flag), PII-minimized sample payloads, reconciliation and DQ rules.
  • reference/connectors/manual_uploads/ - first-class manual/document/file-drop connector with 15 entity templates (budget, forecast, owner report, bid tab, rent survey, comp sheet, approval matrix, property list, staffing model, capex request, draw package, monthly review pack, delinquency report, work order backlog, pm scorecard) plus 16 file templates under file_templates/.
  • reference/connectors/_core/ - shared integration-layer docs: layer_design (raw/normalized/derived), lineage, normalization_patterns, field_mapping_template, raw_to_normalized_design, derived_dependencies, workflow_activation_map (YAML + MD), third_party_manager_oversight, exception_taxonomy, config_overlay_interaction, plus schemas for lineage_manifest, mapping_override_log, benchmark_update_log.
  • reference/connectors/_core/security/ - 15 security and privacy artifacts: security_model, pii_classification, secrets_handling, masking_and_redaction, config_templates, pii_sample_policy, fair_housing_controls, legal_hold_and_retention, audit_trail, unsafe_defaults_registry, approval_gates_for_integration_actions, least_privilege_guidance, config_overlay_interaction, security_testing_guidance.
  • reference/connectors/runbooks/ - 16 operational runbooks covering new source onboarding, schema change, missing file, stale feed, property crosswalk issue, unmapped account, benchmark refresh, failed normalization triage, exception queue review, manual override approval, cutover manual-to-system, connector deprecation, fair housing sensitive flag, financial control gate breach, schema drift escalation, reference rollback.
  • reference/connectors/monitoring/ - alert_policies (25), exception_routing (12 categories), observability_events (22), slo_definitions, alert_channel_design, escalation_matrix.
  • reference/connectors/rollout/ - 5-wave rollout plan, minimum viable data per workflow, pilot property guidance, go-live checklist, rollback plan, success metrics, post-launch monitoring cadence, cutover procedures, pilot-to-production gate.
  • reference/connectors/adapters/ - 8 vendor-family adapter stubs (pms, gl, crm, ap, market_data, construction, hr_payroll, manual_excel) with adapter_manifest schema, lifecycle doc, authoring guide, gotchas, vendor_family_registry.
  • Shared QA library extended under reference/connectors/qa/: freshness_sla, referential_integrity, enum_conformance, temporal_monotonic, provenance_required, cross_source_reconciliation, reasonableness_band.
  • Per-domain depth added to existing 6 connectors (pms, gl, crm, ap, market_data, construction): identity_resolution.md, dq_rules.yaml, reconciliation_rules.md, derived_dependencies.md, field_mapping.yaml. Reconciliation checks expanded from 27 total to 96 total.
  • 7 new pytest modules: test_source_registry.py, test_master_data_crosswalks.py, test_workflow_activation_map.py, test_runbook_structure.py, test_adapter_conformance.py, test_integration_security.py, test_integration_layer_presence.py. Skill-local test count: 69 -> 103. Connector-local test count: 22 -> 89. Total: 91 -> 192.

Changed (residential_multifamily v0.3.0)

  • reference/connectors/_schema/connector_manifest.schema.yaml - connector_kind enum extended from 6 to 8 kinds (adds hr_payroll, manual_uploads). Backward compatible.
  • reference/connectors/_schema/reconciliation_check.schema.yaml - optional template fields added (parameters, severity_default, remediation_pattern, affected_workflows) so shared qa templates validate cleanly.
  • tests/test_connector_contracts.py - REQUIRED_DOMAINS extended to 8 (adds hr_payroll, manual_uploads).
  • hr_payroll/manifest.yaml and manual_uploads/manifest.yaml - updated to declare their true connector_kind now that the canonical enum supports them.

Migration notes (residential_multifamily v0.3.0)

  • Integration-layer work extends reference/connectors/ rather than creating a parallel integrations/ tree. This preserves the existing BOUNDARIES.md layer contract. No canonical _core/ object, metric, workflow, or routing axis was redefined.
  • Directory rename residential_multifamily -> residential-multifamily is still deferred to a dedicated pass.

Added (residential_multifamily v0.2.0 - refinement pass 2026-04-15)

  • overlays/regulatory/ overlay family — regulated affordable housing split into its own family, separate from conventional market-positioning segments. Starter stubs for 6 programs (lihtc, hud_section_8, hud_202_811, usda_rd, state_program, mixed_income) plus 9 shared compliance surfaces (eligibility, income limits, rent limits, UA, recertification, compliance calendar, agency reporting, file audit, escalation sensitivity). No program-specific schedules populated; architecture only.
  • reference/connectors/ vendor-neutral ingestion layer with stubs for 6 source domains (pms, gl, crm, ap, market_data, construction). Each connector: manifest, schema (47 entities across all 6), mapping template, sample input / normalized, reconciliation checks, tests. Reference-layer QA checks under qa/ for record_count, duplicate_id, null_critical_field, date_coverage, unit_count_reconciliation, lease_status_reconciliation, budget_actual_alignment, commitment_change_order_draw.
  • _core/BOUNDARIES.md — canonical boundary document: what lives in core, segment, regulatory, market, org layers. Enforced by tests/test_boundary_rules.py.
  • _core/routing/axes.yaml — new regulatory_program axis with values [none, lihtc, hud_section_8, hud_202_811, usda_rd, state_program, mixed_income]. Default none. Additive; existing packs require no change.
  • _core/routing/rules.yaml — two new rules: r011 (explicit regulatory workflow invocation) and r012 (defensive guard: never auto-load regulatory overlays for conventional segment routes).
  • tailoring/AUDIENCE_MAP.md, tailoring/DIFF_APPROVAL_PREVIEW.md, tailoring/MISSING_DOC_MATRIX.md — canonical tailoring specs.
  • tailoring/question_banks/executive.yaml, finance_reporting.yaml, compliance_risk.yaml, site_ops.yaml — 8-audience split.
  • 4 new pytest modules: test_regulatory_isolation.py, test_connector_contracts.py, test_tailoring_canonical_immutability.py, test_boundary_rules.py.

Changed (residential_multifamily v0.2.0)

  • overlays/segments/ now holds conventional market-positioning only (middle_market + luxury). overlays/segments/affordable/ retained with a deprecation banner and migration pointer to overlays/regulatory/affordable/.
  • _core/routing/axes.yaml — segment values narrowed from [middle_market, affordable, luxury] to [middle_market, luxury]; legacy affordable value accepted for one cycle as a migration alias that prompts the user for regulatory_program and retries routing.
  • _core/schemas/overlay_manifest.yaml — additively extended to support overlay_kind: regulatory and regulatory.program; adds optional parent_overlay, scope.regulatory_program, scope.jurisdiction, and top-level status enum. Pre-existing overlays validate without modification.
  • _core/DESIGN_RULES.md — Rule 6 hierarchy updated to surface regulatory_program as a distinct axis from segment.
  • _core/naming_conventions.md — adds "Directory-slug exception for subsystem roots" (documents the snake_case subsystem-root exception to the repo's kebab-case skill convention) and "Segment vs regulatory program" section.
  • tailoring/SKILL.md — references.reads extended; "Templates used" section updated to enumerate the 8-bank layout.
  • tailoring/question_banks/coo.yaml, cfo.yaml, reporting.yaml — retained with top-of-file deprecation banners pointing at executive.yaml / finance_reporting.yaml.

Deprecated (residential_multifamily)

  • overlays/segments/affordable/ — use overlays/regulatory/affordable/. Old path retained with DEPRECATED.md stub for one refinement cycle to catch late-migrating callers.
  • tailoring/question_banks/coo.yaml, cfo.yaml, reporting.yaml — use executive.yaml and finance_reporting.yaml. Retained for resume of in-flight sessions.

Migration notes (residential_multifamily)

  • Any reference_manifest.yaml reading overlays/segments/affordable/ should update the path to overlays/regulatory/affordable/. Old path continues to resolve for one cycle.
  • Any callsite using axis value segment: affordable should set segment to middle_market or luxury (conventional positioning) and set regulatory_program to the relevant program (or none). The router detects the legacy value and prompts.
  • The residential multifamily subsystem directory is NOT renamed to residential-multifamily in this pass. Tracked for a dedicated ticket.
  • Skill status remains draft; no top-level plugin version bump. The residential_multifamily catalog entry is re-synchronized to status: draft (was stable — mismatch with SKILL.md fixed).

Added

  • residential_multifamily skill — institutional-grade operating system for U.S. residential multifamily property management (534 files, 35,920 lines). Covers the full operating stack:
    • _core/: ontology (30+ objects), 72 metric contracts, routing framework with 3 worked examples, 20-action approval matrix, fair-housing + safety guardrails, alias registry
    • roles/: 15 role packs (property manager flagship; asset manager, TPM-oversight-lead, reporting/finance-ops at same depth; COO/CFO/CEO as rollup consumers)
    • workflows/: 26 workflows with 6 flagships (delinquency_collections, monthly_asset_management_review, bid_leveling_procurement_review, draw_package_review, TPM_scorecard_review, executive_operating_summary_generation)
    • overlays/: segment (middle_market deep + affordable/luxury/high_rise stubs), form_factor x6, lifecycle x6, management_mode x3
    • reference/: 16 category schemas, ~250 starter CSV rows (all tagged sample/starter/placeholder), 16 update-flow walk-throughs
    • templates/: 40 templates with legal-review banners on statutory-notice templates
    • tailoring/: interactive TUI (stdlib + PyYAML, ANSI, session persistence, resume), 7 audience question banks with 124 questions, 23 unit tests
    • tests/: 12 pytest modules, 40/40 pass — enforces no-hardcoded-figures, metric-contract uniqueness, approval gating, fair-housing banners, naming collisions
  • Skill ships with status: draft and self-declared stale-data notice — all reference files are tagged sample | starter | illustrative | placeholder, update before operational use

Changed

  • Skills count: 112 -> 113 (propagated via scripts/catalog-build.py and scripts/catalog-generate.py to README stats table, plugin.json description, registry.yaml, hooks SessionStart prompt, routing table)
  • Prose counts in README, docs/INSTALL.md, docs/install-cowork.md, docs/install-desktop.md, docs/WHAT-TO-USE-WHEN.md, docs/install-guide.md updated to 113 (historical release notes under docs/releases/ left unchanged)

Added (CI)

  • scripts/prose-drift-check.py — grep-based CI guard that flags hardcoded skill/agent/calculator/reference-file counts in prose outside the CATALOG:STATS markers. Wired into .github/workflows/ci.yml so future skill additions fail CI until every current-state doc is updated or explicitly wrapped in <!-- PROSE-DRIFT:IGNORE-START --> / <!-- PROSE-DRIFT:IGNORE-END --> markers. Closes the gap the generator never covered (14 stale "112" references in prose surfaced during this merge).

Notes

  • No version bump in this entry — the skill status is draft, version should bump when the subsystem moves to deployed

[4.1.2] - 2026-04-13

Added

  • Install docs for Codex, Gemini, Grok, Manus targets
  • Cosign verification steps per target
  • Non-technical quickstart section in docs/INSTALL.md

Changed

  • Collapsed checksum/signing details into developer-only details blocks

Notes

  • Doc-only release; no plugin behavior changes

[4.1.1] - 2026-04-02

Added

  • Desktop and portable zip build targets in release.yml
  • Source archives attached to each release
  • Cosign signing for all release artifacts
  • Consolidated SHA256SUMS across targets

Changed

  • Release workflow produces 4 platform targets per tag (macOS DMG, Windows EXE, desktop zip, portable zip)

[4.1.0] - 2026-04-02

Added

  • Same-repo marketplace install path
  • Cowork install guidance (Desktop zip upload flow)
  • Structured installer telemetry (install start, success, failure, edge cases)

Fixed

  • /doctor now recognizes the plugin (plugin.json moved to repo root)
  • Windows installer verification and plugin cache layout
  • PS1 installer: step-by-step progress and structured error reporting

[4.0.0] - 2026-04-02

Added

  • MCP server (src/mcp-server.mjs): zero-dependency stdio JSON-RPC server with 8 tools (cre_route, cre_list_skills, cre_skill_detail, cre_workspace_create/get/list/update, cre_send_feedback) for Claude Desktop visibility
  • Feedback retry outbox (src/hooks/feedback-outbox.mjs): failed remote submissions are queued in ~/.cre-skills/outbox.jsonl and retried on next session start (4s timeout per request, max 5 attempts before eviction)
  • Canonical catalog system: src/catalog/catalog.yaml as single source of truth for all plugin metadata (196 items: 105 skills, 54 agents, 9 commands, 12 calculators, 10 orchestrators, 6 workflows)
  • Catalog schema: src/catalog/catalog.schema.json
  • Build script: scripts/catalog-build.py (scan repo -> catalog.yaml + dist/catalog.json)
  • Generator script: scripts/catalog-generate.py (catalog -> README, hooks, plugin.json, routing, registry)
  • Catalog-driven router: src/routing/skill-dispatcher.mjs now reads dist/catalog.json with artifact-aware matching, confidence scoring, downstream recommendations, and markdown fallback
  • Output styles: src/templates/output-styles/ with 5 format templates (exec-brief, ic-memo, pm-action-list, lender-brief, lp-update)
  • userConfig in plugin.json: primary_asset_types, preferred_markets, default_output_style, feedback_mode, brand_name
  • ADR: docs/adr/0001-catalog-source-of-truth.md
  • Migration guide: docs/MIGRATION.md
  • Release checklist: docs/release-checklist.md
  • Workspace entrypoints: 7 top-level skills (deal-intake, lease-strategy-papering, asset-ops-cockpit, capital-projects-development, fund-lp-reporting, navigator, plugin-admin)
  • Persistent workspace state: local JSON storage for session continuity
  • Next-best-action footer: mandatory structured output for workspace skills

Changed

  • Build system refactor: all source content (skills, agents, commands, hooks, routing, orchestrators, calculators, lib, catalog, MCP server, templates) moved from repo root into src/. Symlinks removed. scripts/, docs/, tests/, dist/, registry.yaml, and README.md remain at repo root. Build output goes to builds/.
  • Feedback default mode: ask_each_time -> local_only (privacy-first)
  • Feedback backend_url default: pre-configured -> empty (opt-in remote submission)
  • registry.yaml: manually maintained -> generated from catalog
  • Router: markdown-table parsing -> catalog-driven with fallback
  • README Key Stats: hardcoded -> generated from catalog (18 categories)
  • Plugin version: 3.0.0 -> 4.0.0
  • Telemetry now enabled by default (opt-out model). Records only skill slug + date + anonymous UUID. No deal data, financials, prompts, or identity. All data stays local at ~/.cre-skills/telemetry.jsonl
  • First-run notice clearly explains what is and is not tracked, and how to opt out
  • Windows installer registers in ~/.claude/ plugin system instead of %APPDATA%\Claude\skills\
  • Stale doc references: added notices to USER-GUIDE.md for unimplemented CLI scripts

Fixed

  • PRIVACY.md: claimed remote submission was "Future -- Not Available Yet" when code had it active by default. Now accurately describes it as available but opt-in.
  • docs/feedback-system.md: contradicted itself on default feedback mode
  • feedback-summary command: now reads both feedback.jsonl and feedback-log.jsonl
  • Hooks.json SessionStart prompt: counts now generated from catalog
  • Windows Desktop not seeing installed plugin (wrong install directory)

[3.0.0] - 2026-04-01

Added

  • Feedback system: /cre-skills:send-feedback and /cre-skills:report-problem commands for structured feedback intake without leaving the session
  • Remote feedback submission via Vercel Function + Supabase backend at cre-skills-feedback-api.vercel.app (default mode: ask_each_time, user confirms each send)
  • Automatic redaction of file paths, emails, digit sequences, and env vars from feedback text before storage (scripts/redact-feedback.mjs, 38 behavioral tests)
  • Weekly feedback prompt: non-intrusive reminder at session end after 7+ days of use, shown once per 7 days, only when CRE skills were used in the session
  • Feedback schemas: feedback-submission.schema.json, feedback-config.schema.json
  • Architecture documentation: docs/feedback-system.md
  • 38 redaction tests (tests/test-redaction.mjs) + 4 new structural integrity tests
  • firstRunAt timestamp tracking for install age calculation

Changed

  • Total commands: 7 -> 9 (added send-feedback, report-problem)
  • Config model: added feedback block (mode, include_context, backend_url, last_prompt_shown_at)
  • hooks/hooks.json: SessionStart prompt updated with new commands
  • hooks/telemetry-init.mjs: feedback config initialization + backfill for existing installs
  • hooks/session-summary.mjs: weekly feedback prompt logic
  • PRIVACY.md: documented structured feedback data flow, redaction rules, remote submission
  • README.md: feedback section, updated stats (9 commands), project structure

[2.5.0] - 2026-03-28

Added

  • construction-cost-estimator: full CSI MasterFormat cost estimation with 11 workflows including visual input (floor plans/drawings), generative output (ASCII mockups), and co-creation design loops. Includes construction_estimator.py calculator (zero dependencies), 925-line CSI cost database, 60+ city regional cost factors, and soft cost benchmarks reference.
  • property-management-orchestrator: deep PM command center with 10 workflows across all asset types. Routes to downstream skills (building-systems, work-order-triage, PM operations, etc.). Includes 8 asset-type module stubs (multifamily through hospitality) for v3.0 deep-dive, IREM/BOMA KPI benchmarks, monthly report template, and vendor management framework.
  • space-planning-redesign-orchestrator: multi-agent orchestrator for space redesign ideation with 5 specialist subagents (space programmer, design visualizer, cost estimator, market validator, ROI analyst). Includes amenity cost benchmarks, tenant survey templates, competitive audit framework, and ROI model template. Status: stub (full orchestration in v3.0).
  • Windows .exe installer via Inno Setup (no admin privileges, auto-detects Claude Code + Desktop)
  • GitHub Actions workflow for building Windows installer on windows-latest runner
  • 12th Python calculator: construction_estimator.py

Changed

  • Total skills: 99 -> 105 (6 new)
  • Total reference files: 225 -> 247 (22 new)
  • routing/CRE-ROUTING.md updated with 3 new skill routing triggers
  • hooks/hooks.json SessionStart prompt updated to 102 skills, 55 agents
  • Install.ps1 added for Windows post-install configuration

[2.0.0] - 2026-03-25

Added

  • 19 new skills (total: 99 skills): title-commitment-reviewer, tenant-credit-analyzer, term-sheet-builder, loan-document-reviewer, transfer-document-preparer, funds-flow-calculator, lease-option-structurer, lease-trade-out-analyzer, gp-performance-evaluator, fund-terms-comparator, lp-data-request-generator, sec-reg-d-compliance, monte-carlo-return-simulator, property-management-operations, distribution-notice-generator, 1031-pipeline-manager, deal-attribution-tracker, emerging-manager-evaluator, fund-raise-negotiation-engine
  • 11 Python calculator scripts (zero dependencies): debt_sizing, covenant_tester, npv_trade_out, option_valuation, proration_calculator, quick_screen, tenant_credit_scorer, transfer_tax, waterfall_calculator, monte_carlo_simulator, fund_fee_modeler
  • Orchestration engine: 10 multi-agent pipelines (acquisition, capital-stack, hold-period, disposition, development, fund-management, research, strategy, portfolio, lp-intelligence) with 10 JSON configs, 16 prompt files, 8 investor profiles, strategy matrix, challenge layer, disagreement/reversal schemas, and cross-chain handoff registry
  • 14 lifecycle agents across research, strategy, asset-management, portfolio, fund, disposition, and LP intelligence categories (total: 54 agents)
  • 225 reference files (up from 171 in v1.0.0)
  • Brand guidelines system: auto-loads from ~/.cre-skills/brand-guidelines.json for 8 deliverable skills
  • PostToolUse hook (telemetry-capture.mjs) -- tracks skill invocations when telemetry is opt-in
  • Stop hook (session-summary.mjs) -- writes session_end record and optional feedback prompt
  • telemetry-init.mjs SessionStart hook -- initializes ~/.cre-skills/config.json on first run
  • 4 new commands: brand-config, orchestrate, usage-stats, feedback-summary (total: 7 commands)
  • scripts/update.sh -- pulls latest, handles major version migration, preserves user data
  • scripts/uninstall.sh -- clean removal with user data preservation prompt
  • scripts/verify-install.sh -- full health check (skills, refs, calculators, hooks, data dir)
  • docs/install-guide.md -- comprehensive installation guide for all platforms
  • PRIVACY.md, SECURITY.md

Changed

  • License changed from MIT to Apache 2.0 (adds explicit patent grant)
  • NOTICE file added for Apache 2.0 compliance with Avi Hacker attribution
  • scripts/install.sh updated for v2: Node.js/Python checks, v1->v2 migration, backup, calculators
  • Install.command version string updated to v2.0.0
  • README.md rewritten for v2.0.0 with full skill catalog, workflow chains, calculators, brand guidelines
  • hooks/telemetry-capture.mjs: replaced /dev/stdin with fd 0 for Windows WSL compatibility
  • verify-install.sh: counts all reference files (*.md + *.yaml), not just *.md

Breaking Changes

  • License: MIT -> Apache 2.0. Existing installations continue to function; license governs redistribution and use going forward.

[1.0.0] - 2026-03-17

Added

  • 80 CRE skill packs across 16 subcategories
  • 171 reference files (formulas, templates, frameworks, worked examples)
  • 40 expert subagent definitions across 8 categories
  • 6 workflow chain documents
  • 3 plugin commands (cre-route, cre-workflows, cre-agents)
  • SessionStart hook for context-efficient routing
  • Machine-readable skill registry (registry.yaml)
  • CRE routing index for skill discovery
  • MIT license